Resource Allocation for Malaria Prevention

1
Resource Allocation for Malaria Prevention
World Health Organization
Final PresentationApril 17, 2008
Michele Cataldi Christina Cho Cesar Gutierrez
Jeff Hull Phillip Kim Andrew Park
Sponsor Contact Jason Pickering, PhD. Faculty
Advisor Julie Swann, PhD.

• Disclaimer This document has been created in the
  framework of a senior design project. The
  Georgia Institute of Technology does not
  officially sanction its content.

2
Agenda

• Client Background
• Problem Description
• Solution Strategy
• Model
• Deliverables
• Value

3
Client Background

• World Health Organization
• Responsible for providing leadership to all UN
  member nations on global health matters
• Public Health Mapping Group
• Data analysis, process and visualization via
  Geographic Information Systems (GIS)

4
Problem Description

• Malaria
• 300-500 million cases per year and over 1 million
  deaths
• Prevention methods
• Indoor Residual Spraying (IRS)
• Long-Lasting Insecticide Treated Bed Nets (LLIN)
• No existing procedure for optimal allocation of
  limited prevention resources
• Arbitrary distribution
• Detrimental effects of excessive spraying

5
Solution Strategy

• Create a systems-based approach to minimize the
  incidence of malaria with limited resources.
• Swaziland as pilot country
• Historical data availability
• Wide range of conditions

http//www.mara.org.za/
http//en.wikipedia.org/wiki/Swaziland
6
Data Sources

• Mapping Malaria Risk in Africa (MARA)
• Percentage risk estimation by region
• 5x5 km spatial resolution
• Start and end months of high malaria transmission

http//www.mara.org.za/
http//www.mara.org.za/
7
Data Sources

• HealthMapper

Road Infrastructure
Facility Infrastructure
8
Data Sources

• Costs and other intervention data
• World Health Organization Malaria Costing Tool
• UN Millennium Project

9
Model Objective

• An optimization model will allow for a
  systems-based approach to resource allocation and
  deployment for malaria prevention.
• Decisions include
• Where to locate Distribution Centers (DCs)
• How many DCs to open
• When DCs should be open
• What regions DCs should serve
• When to cover each zone
• Number of people to protect in each zone
• Labor, trucks, equipment, insecticide/nets to
  base at DCs
• Labor, trucks, equipment to allocate to each zone

10
Model Overview
DC Placement Heuristic
Zone Assignment Heuristic
Resource Deployment Model
Decision Tool
11
DC Placement Heuristic

• Potential locations for DCs
• Factors considered
• Population
• Malaria risk
• Infrastructure
• Scalable for other countries
• Distance constraints adjusted by estimated area

Max. distance from center point
Min. distance between DCs
where d represents ½ the (estimated) height of
the country, and n the number of DCs
12
DC Placement Heuristic
Swaziland 5 DC Placement
MalariaRisk
Population
13
Zone Assignment Heuristic

• Customer zones are serviced by a single DC
• Straight-line distance DC to customer zone
• Road factor of zone considered (paved, unpaved)

Zone Assignment with 3 DCs
Zone Assignment with 5 DCs
14
IRS Resource Deployment Model

• Objective Maximize the number of people
  protected by a prevention method who are at risk
  of malaria.
• Output scheduled deployment plan
• What zones to target for spraying
• When to deploy in each zone
• How many people in each zone to protect
• Resources to base at DCs

15
Assumptions

• MARA
• Risk and transmission season accurately
  represented by MARA
• 5x5 km MARA grids aggregated into 15x15 km zones
• Intervention
• IRS with DDT
• Materials ordered once per year, prior to
  deployment
• 1 spray cycle per year
• Straight line distances adjusted for road
  conditions of zone

Distribution Center
Zone(s)
Zone(s)
Zone(s)
Zone(s)


time 1
time 2
time 3
time T
16
IRS Constraints

• Deployment restricted by

Capacity Relative Effectiveness Costs/Budget
Truck capacity DC capacity Amount of resources based at DCs Zone population Duration of DDT effectiveness Concentration of DDT per m2 Coverage rate of spray personnel Labor wages Opening and operating DCs Vehicle rental and travel costs Equipment purchase and repair Cost of DDT
17
LLIN Resource Deployment Model
Advertisement of net pickup place and time to
zones
DCs open for net pickup and instruction on proper
use
time(0) ... time (DC open)
time(DC open) ... time (DC close)
Zone(s)
Zone(s)
Zone(s)
Zone(s)
DC
DC
Zone(s)
Zone(s)
Zone(s)
Zone(s)

• Adapted output
• When to open the DCs
• What zones to target
• Number of public health workers and supervisors
  at DCs
• Extent of advertising in targeted zones

18
Recommendation
19
Recommendation

• Labor based at each DC

DC Zone Labor
1 136 35
2 219 35
3 435 50
4 537 50
5 - -
6 - -
20
Recommendation

• Deployment schedule

For full deployment schedule, see animation
21
Sensitivity Analysis
Parameter Spray rate per worker Parameter Spray rate per worker      
(houses/day)        
Factor 0.1 0.9 1.1 1.9
         
Objective Value 8,642,903 40,212,463 42,917,104 46,045,932
Objective Value / Total Cost 17.31 80.42 86.14 92.72
         
? from base -0.795 -0.046 0.022 0.100
         
Parameter Value 0.77 6.93 8.47 14.63
? objective / ? Parameter 0.88 0.46 0.22 0.11
22
Model Interface

• Decision-making application using Excel and VBA
• Linked to Xpress-MP

23
Deliverables

• Optimization model
• Description, specification of model
• Decision-making tool
• Test interface in Excel
• Output
• Sensitivity analysis
• Objective response to changes in parameters
• Documentation
• All assumptions, processes, and methodology

24
Value

• Use of heuristics to estimate expected current
  deployment behavior
• 3 heuristic variations, prioritize zones to cover
  by
• Greatest risk first
• Greatest population first
• Greatest combined risk and population first
• All variations assume
• 1 DC in Mbabane (capital)
• Equivalent objective, budget, and resource
  constraints

25
Value
/Person Covered/Year Cost Reduction in Model
Model 1.32 -
Heuristic 1 2.19 -39.73
Heuristic 2 2.52 -47.62
Heuristic 3 2.58 -48.84
Research Average 2.59 -49.03
The American Society of Tropical Medicine and
Hygiene, http//www.ajtmh.org/cgi/reprint/77/6_Sup
pl/138
26
Value
Effective Coverage Coverage Increase in Model
Model 376,874 -
Heuristic 1 213,087 76.86
Heuristic 2 187,070 101.46
Heuristic 3 191,525 96.78
27
Value
of People (millions) of at Risk Population
Total at Risk in Africa 672 -
Current Coverage 193.05 28.73
Potential Coverage 378.79 56.37
Africa alone loses an average of 12 billion US
dollars of income per year, because of malaria.
WHO/Gates Foundation Project
Malaria Atlas Project http//www.map.ox.ac.uk
28
Summary

• Problem Description
• Solution Strategy
• Model
• Recommendations
• Value

29

• Questions?